Documente Academic
Documente Profesional
Documente Cultură
Services
Ian Ku, You Lu, Mario Gerla Francesco Ongaro
Department of Computer Science Department of Computer Science and Engineering
University of California Los Angeles University of Bologna
Los Angeles, USA Bologna, Italy
{ianku, youlu, gerla} @cs.ucla.edu francesco.ongaro@studio.unibo.it
Abstract Vehicular Ad Hoc Networks (VANETs) have in recent VANETs are now a reality and support a variety of new
years been viewed as one of the enabling technologies to provide a services and protocols. However, there are still challenges in
wide variety of services, such as vehicle road safety, enhanced the deployment of VANETs applications, such as unbalanced
traffic and travel efficiency, and convenience and comfort for flow traffic among multi-path topology, and inefficient network
passengers and drivers. However, current VANET architectures utilization. Therefore, open and flexible vehicular architectures
lack in flexibility and make the deployment of services/protocols are key requirements to allow experimenters to test their
in large-scale a hard task. In this paper, we demonstrate how solutions in productive environments, as well as to improve the
Software-Defined Networking (SDN), an emerging network management of network resources, applications, and users.
paradigm, can be used to provide the flexibility and
programmability to networks and introduces new services and To address these challenges, we look at Software-
features to todays VANETs. We take the concept of SDN, which Defined Networking (SDN) [3]. Nowadays, SDN has emerged
has mainly been designed for wired infrastructures, especially in as a flexible way to control the network in a systematic way,
the data center space, and propose SDN-based VANET with OpenFlow [4, 5] as the most commonly used SDN
architecture and its operational mode to adapt SDN to VANET protocol. The flexibility of SDN makes it an attractive
environments. We also discuss benefits of a Software-Defined approach that can be used to satisfy the requirements of
VANET and the services that can be provided. We demonstrate VANET scenarios. Applying SDN principles to VANETs will
in simulation the feasibility of a Software-Defined VANET by bring the programmability and flexibility that is lacking in
comparing SDN-based routing with traditional MANET/VANET
todays distributed wireless substrate, while simplifying
routing protocols. We also show in simulation fallback
mechanisms that must be provided to apply the SDN concept into
network management and enabling new V2V and V2I services.
mobile wireless scenarios, and demonstrate one of the possible While most SDN/OpenFlow based advances has been
services that can be provided by a Software-Defined VANET. focused on wired networks, especially for datacenters, increase
attention has been made on incorporating SDN/OpenFlow into
Keywords-Software-Defined Networking; VANET; wireless the wireless domain [6]. While its scope has been primarily
networks; focused on carrier backbones and access networks, the usage of
I. INTRODUCTION SDN in other wireless scenarios has been gaining attention
from both academia and industry. OpenRoads [7] envisions
Vehicular Ad Hoc Networks (VANETs) have received a lot that users will move between wireless infrastructures.
of interest in recent years. It is an active area of research CloudMAC [8] proposes virtualized access points. The
because of its potential to improve vehicle road safety, enhance Wireless & Mobile Working Group (WMWG) [9] in ONF
traffic and travel efficiency, and provide convenience and focuses on wireless backhaul, cellular Evolved Packet Core
comfort for passengers and drivers. One example would be the (EPC), and unified access and management across enterprise
Intelligent Transport Services (ITS) [1]. wireless and fixed networks (e.g., campus Wi-Fi).
With the expected growth in mobile devices and mobile Other works on wireless SDN include OpenFlow in
traffic, Vehicle-to-vehicle (V2V) and Vehicle-to-Infrastructure wireless mesh environments [10], OpenFlow in smartphone as
(V2I) communications are expected to become more in demand an application [11], OpenFlow in wireless sensor networks
and will continue to grow [2]. VANETs can be used to provide [12], SDN in heterogeneous networked environments [13], and
a wide range of services, including both safety and non-safety SDN for handover management in heterogeneous networks
related applications. Examples include vehicular safety [14]. However, it is necessary to understand and extend the
traffic management services, surveillance services, and mobile usage of SDN/OpenFlow in VANETs.
vehicular cloud services.
In this paper, we focus on applying SDN into VANETs. In forwarding while the other is exploited for the network traffic
specific, we look at the architecture, operations, and benefits of control.
Software-Defined VANET services and new functionalities to
support them. By decoupling the control and data planes in OpenFlow [4, 5] is the most commonly used protocol for
VANETs, network intelligence and state can be logically communication between the SDN control plane and data plane.
centralized and the underlying network infrastructure is Fig. 2 shows a high level concept of SDN. In this paper,
abstracted from the applications. Thus, it will be possible to OpenFlow is used as base and is integrated to VANET wireless
have highly adaptive, flexible, programmable, and scalable environments.
VANETs environments. A use-case on routing is presented to
demonstrate the benefits of integrated SDN VANETs
architectures in forwarding data in multipath scenarios.
The reminder of the paper is structured as follows. Section
II provides background information on VANET and
SDN/OpenFlow. Section III describes the architecture and
operations of a Software-Defined VANET. Benefits and
services for Software-Defined VANETs are presented in
Section IV. Section V presents simulation evaluation and
Section VI concludes the paper.
II. BACKGROUND AND TERMINOLOGIES
Figure 2. Software Defined Networking Concept
In this section we describe some background information
and terminologies on VANET and SDN/OpenFlow used Fig. 3 shows an OpenFlow network comprising the two
through the paper. main components: the OpenFlow controller and several
OpenFlow-enabled switches that communicate using a secure
A. VANET channel.
In a typical VANET, Vehicles communicate with each The controller is a software program element that is used to
other through V2V communication in Ad hoc fashion, and V2I modify the content of flow tables, which are located in each
communication through road-side-units (RSU) and mobile OpenFlow-enabled switch. Each flow table contains a list of
broadband (e.g. 4G/LTE). Fig. 1 shows the components and flow entries, which consist of Match Fields, Priority, Counters,
communications with a typical VANET. Instructions, Timeouts, Cookie, and actions associated. When a
packet comes to the switch there is a lookup into the matching
Traditional VANET services include vehicle and road field of each entry. If there is a match, the packet will be
safety services, traffic efficiency and management services, and processed according to the actions of the matching flow entry.
infotainment services. Vehicle and road safety services are On the other hand, if a packet does not match, a table-miss
those that target the decrease of traffic accidents and loss of life occurs. In this case, different actions can be taken as specified
to vehicle occupants. Traffic efficiency and management in the table-miss flow entry, for instance, the OpenFlow-
services aim to improve traffic flow, traffic coordination, and enabled switch can encapsulate the packet and send it to the
to provide local and map information. Infotainment services controller through the secure channel or directly drop the
aims to provide information and entertainment such as packet.
multimedia data transfer and global Internet access.
OpenFlow controller controls the behavior of the network
B. SDN/OpenFlow by sending flowmods packets to modify the content of flow
The core concept of SDN [3] is the separation between the tables. The controller has two ways to add rules in the switch:
control plane and the data plane. The latter is used for the data (I) proactively, where the controller takes the initiative and
adds rules before packet arrival into the network; or (II) Long range wireless connection i.e., LTE/Wimax for control
reactively, where the controller reacts because of an in the plane, and high bandwidth wireless connection i.e., Wi-Fi for
network, such as a previously unrecognized packet. An data plane. The practical reason is that in VANETs not all
example of such an operation would be an Access Point nodes are easily reachable from the Infrastructure via RSUs.
(AP)/switch that sends a data packet to the controller (by Fig. 4 shows the communication between components in our
encapsulating the packet in an OpenFlow control packet called Software-Defined VANET.
packet-in ) because it does not know how to deal with it. Then,
the controller sends the flowmods packet to the APs/switches
with instructions.
One notable feature in an OpenFlow network is that once a
specific traffic flow matches the flow table, the OpenFlow-
enabled switch knows how to treat this flow and does not
need further interactions with the OpenFlow controller. While
this allows switches to forward traffic efficiently, issues arise
when the flow table rules are no longer consistent with the
network condition. In other words, if network conditions such
as topology have changed, until the controller inserts/updates
the flow table entry, an OpenFlow-enabled switch will use the
old (and potentially incorrect) rule. Section V shows more
details about this issue.
In mobile networks, the introduction of SDN and
OpenFlow will enable the programing of base stations' wireless Figure 4. Software-Defined VANET Communications
data plane and enhance the functionalities of the core networks. Fig. 5 shows the internal components of a SDN wireless
Thus, it will improve the management of resources and mobile node. It contains all the functionality of an OpenFlow-enabled
devices and create a great opportunity for new services and switch in traditional OpenFlow networks, plus additional
control functions. In dynamic wireless mobile environments, intelligence to enable different modes of operation in VANET
such as VANETs, the use of SDN can reduce interference; environments. The number of WiFi interfaces used as data
improve the usage of channels and wireless resources, as well channel is based on configuration and the service that the SDN
as the routing of data in multi-hop and multi-path scenarios. wireless node is required to support. The SDN module is the
We describe the benefits of Software-Defined VANETs later in combination of packet processing and the interface that accepts
Section IV. input from a separated control plane.
III. SOFTWARE DEFINED VANET
In this section, we describe the architecture of a Software-
Defined VANET and its operations. The goal is to describe
how VANETs take advantage of SDN concepts and
functionalities to improve resource utilization, select best
routes, and facilitate network programmability.
A. Architecture Overview
To enable a SDN based VANET system, our architecture
incorporates the following SDN components:
Figure 5. SDN wireless node internals
SDN controller: The logical central intelligence of the
SDN based VANET system. The SDN controller Each SDN wireless node has a local SDN agent, which
controls the network behavior of the entire system. functionality depends on what features are enabled on the SDN
wireless node. This local SDN agent can either be the backup
SDN wireless node: The data plane elements that are controller when connection to the SDN controller is lost or the
controllable by the SDN controller. They are the primary SDN intelligence while receiving input from the SDN
vehicles that receive control message from the SDN controller. Traditional Ad hoc routing protocols (e.g., GPSR,
controller to perform actions. AODV, DSDV, and OLSR) are supported in agents as fallback
SDN RSU: Stationary data plane elements that are mechanisms, to allow the SDN network to revert back to Ad
controllable by the SDN controller. They are the hoc network operation even in the case where SDN controller
infrastructure RSUs that are deployed along road communication is unavailable. In scenarios where the
segments. connection to a SDN controller is stable and has full control,
this SDN agent has minimal intelligence.
Our proposed architecture extends SDN to operate in
mobile wireless VANET scenarios. In our architecture, we One distinct characteristic of Ad hoc networks is that the
choose to use different wireless technologies for controlling nodes act both as Hosts (sending/receiving traffic) and Routers
and forwarding planes as expected in future VANET systems: (forwarding traffic on behalf of other nodes). An SDN wireless
node is therefore both an SDN data plane forwarding element controller does not hold complete control, but instead
and an end-point for data. Traffic from any wireless node (e.g. can delegate control of packet processing details to
application traffic) will run through its own SDN module local agents. Therefore control traffic is exchanged
before being sent, which allows the SDN controller to between all SDN elements. One example would be that
determine the access of user traffic into the network. instead of sending complete flow rules, the SDN
controller instead sends out policy rules which define
B. Operations Overview general behavior, while the SDN wireless nodes and
While the concept of SDN is the separation of control and SDN RSUs use local intelligence for packet forwarding
data plane, there are differences in how a Software-Defined and flow level processing. In specific, the SDN
VANET can operate based on the degree of control of the controller instructs SDN wireless nodes and RSUs to
SDN controller. We classify our architecture into three run a specific routing protocol with certain parameters.
operational modes:
Central Control Mode: This is the mode where the
SDN controller controls all the actions of underlying
SDN wireless nodes and RSUs. In specific, all the
actions that the SDN data element performs are
explicitly defined by the SDN controller. As shown in
Fig 6, the SDN controller will push down all the flow
rules on how to treat traffic.
Figure 9. Grid Road Network We can see that Packet delivery ratio immediately starts to
drop as SDN controller no longer insert fresh rules for SDN
Fig 10 shows the comparison of SDN-based VANET
wireless nodes. This demonstrates that operating a Software-
routing to other traditional MANET/VANET routing protocols,
Defined VANET in central control mode is dangerous if SDN
including GPSR, OLSR, AODV, and DSDV. We use this
controller communication is not reliable enough. The nature of
evaluation to demonstrate the feasibility of a Software-Defined
a VANET is that nodes will move around quickly, and stale
VANET.
rules will become obsolete much more quickly compared to a
scenario where node mobility is low.
We then show the same scenario, except that this time the
fallback mechanism, running GPSR routing, is triggered when
SDN controller communication is lost. Fig. 12 shows the
packet delivery ratio of this scenario.
VANET, and we described several different operational modes
and the services that can be provided. We demonstrate in
simulation several points: (I) the feasibility of a Software-
Defined VANET by comparing SDN-based routing with
traditional MANET/VANET routing protocols, (II) how
fallback mechanism is a key feature that must be provided to
apply the SDN concept into mobile wireless scenarios, and (III)
transmission power adjustment as one of the possible services
that can be provided by Software-Defined VANET.
For future work, there are several directions that we intend
to explore. First, although we demonstrated how a fallback
mechanism can maintain good packet delivery in the case of
SDN controller failure, there are more advanced scenarios that
Figure 12. SDN controller failure with GPSR as fallback should be considered. For example, there are cases of partial
We can see that after an initial drop (after the loss of SDN controller connectivity loss, where only a subset of
communication with SDN controller and before GPSR is mobile nodes loses communication to the controller. In this
activated), good packet delivery ratio is restored. Once case, the isolated nodes might need to form their own SDN
communication with SDN controller is restored, the system cluster, or nodes with connectivity to the SDN controller can
once again reverts back to SDN-based routing. act as relay nodes to relay rules. The best action to take will
depend on several factors such as node density, mobility
C. Adaptive Transmission Range Selection pattern, or others.
In this evaluation, we demonstrate how allowing the SDN
Second, although we demonstrated the feasibility of a
controller to dynamically control the transmission power of
Software-defined VANET, our current architecture still
SDN wireless nodes can improve packet delivery. The requires infrastructure support (e.g. LTE). Therefore, we would
simulation again is performed over the SUMO generated grid
like to investigate alternate Software-Defined VANET
road network; however the node density is 30 nodes. Fig. 13
architectures such as where SDN controller transmits control
shows the result, with the dash line marking the time when the
traffic in P2P mode using WiFi channels. While this allows to
SDN controller raises the transmission power so that
build a wireless SDN system that is completely distributed and
transmission range is now approximately 400 meters (up from
thus does not need infrastructure support, the communication
the previous 250 meters).
with the SDN controller can be delayed and even interrupted
causing new complications.
Third, we intend to investigate on additional Software-
Defined VANET services and their performances. Such as the
safety and surveillance services that we proposed, and also
other new services that can be deployed by taking advantages
of the benefits of SDN.
ACKNOWLEDGMENT
This work was partially supported by CNPq, CAPES, and
FAPESP (Grant 2012/04945-7).
REFERENCES
Figure 13. Adaptive Transmission Range Selection
[1] M. Gerla. Vehicular Cloud Computing, VCA 2012 Proceedings,
We can see how immediately how packet delivery ratio Cyprus, June 2012
becomes higher, as the new transmission range provides much [2] C. Harsch, A. Festag, and P. Papadimitratos. Secure position-based
better connectivity between nodes. The SDN controller can routing for VANETs. In Proceedings of IEEE 66th vehicular
technology conference (VTC-2007), Fall 2007 (pp. 2630), September
make the judgment to adjust transmission power because it has 2007
global information on the entire VANET scenario. In specific, [3] N. Mckeown, Software-defined networking. INFOCOM keynote talk,
in our evaluation the SDN controller increases transmission Apr, 2009
power because based on the information gathered from the [4] N. McKeown, T. Anderson, H. Balakrishnan, G. Parulkar, L. Peterson, J.
SDN wireless nodes, it determines that the connectivity Rexford, S. Shenker, and J. Turner, Openflow: enabling innovation in
between SDN wireless nodes is too low. campus networks, SIGCOMM Comput. Commun. Rev., vol. 38, no. 2,
pp. 6974, 2008.
VI. CONCLUSION AND FUTURE WORK [5] OpenFlow Switch Specification, Version 1.4.0 (Wire Protocol 0x05).
[Online]. Available: https://www.opennetworking.org/images/stories/
In this paper, we propose the architecture and services downloads/sdn-resources/onf-specifications/openflow/openflow-spec-
toward a Software-Defined VANET. The architecture captures v1.4.0.pdf
the components and requirements needed to deploy SDN in
[6] ONF Solution Brief, OpenFlow-Enabled Mobile and Wireless [13] M Mendonca, K Obraczka, and T Turletti, The Case for Software
Networks, September 2013 Defined Networking in Heterogeneous Networked Environments, ACM
[7] K.-K. Yap, M. Kobayashi, R. Sherwood, T.-Y. Huang, M. Chan, N. converence on CoNEXT student workshop, 2012.
Handigol, and N. McKeown, OpenRoads: empowering research in [14] C. Guimares, D. Corujo, R. L. Aguiar, F. Silva, and P. Rosa,
mobile networks, SIGCOMM Comput. Commun. Rev., vol. 40, no. 1, "Empowering Software Defined Wireless Networks Through Media
pp. 125126, January 2010. Independent Handover Management", Proc. 2013 IEEE Globecom,
[8] J. Vestin, P. Dely, A. Kassler, N. Bayer, H. Einsiedler, and C. Peylo, Atlanta, USA, Dec 2013
CloudMAC: towards software defined WLANs, ACM SIGMOBILE [15] P. Pawelczak, R. Venkatesha Prasad, L. Xia, and I. Niemegeers.
Mobile Computing and Communications Review, vol. 16, no. 4, pp. 42 Cognitive radio emergency networks-requirements and design. In the
45, 2013. First IEEE International Symposium on New Frontiers in Dynamic
[9] Wireless & Mobile Working Group (WMWG). [Online]. Available: Spectrum Access Networks (DySPAN), pages 601{606. IEEE, 2005.
https://www.opennetworking.org/images/stories/downloads/working- [16] I. Akyildiz, W. Lee, and K. Chowdhury. Crahns: Cognitive radio ad
groups/charter-wireless-mobile.pdf hoc networks. Ad Hoc Networks, 7(5):810{836, 2009.
[10] P. Dely, A. Kassler, and N. Bayer. Openflow for wireless mesh [17] NS-3, http://www.nsnam.org/
networks. In Proceedings of 20th International Conference on [18] Simulation of Urban Mobility (SUMO), http://sumo-sim.org/
Computer Communications and Networks (ICCCN), pages 1{6. IEEE,
2011. [19] J. Broch, D. Maltz, D. Johnson, Y.-C. Hu, and J. Jetcheva. A
Performance Comparison of Multi-Hop Wireless Ad Hoc Network
[11] P. Baskett, Y. Shang, W. Zeng, and B. Guttersohn. SDNAN: Software- Routing Protocols. In Proceedings of the Fourth Annual ACM/IEEE
Defined Networking in Ad hoc Networks of Smartphones, Consumer International Conference on Mobile Computing and Networking
Communications and Networking Conference (CCNC), 2013 IEEE (MobiCom98), pages 8597, Dallas, TX, October 1998. ACM
[12] T. Luo, H. Tan, and T. Quek, Sensor openflow: Enabling sofware-
defined wireless sensor networks, Communications Letters, IEEE, Vol.
16 , Issue: 11, 2012